4 Limitations of Facial Recognition Technology

A limitation of facial recognition technology for the general public. Accustomed to seeing computers on television and movies comparing computers to motor vehicles and other databases in seconds, many private citizens are surprised that our country’s top law enforcement agencies are not as technologically advanced as those in Boston. As a result, perhaps the country’s security is vulnerable to it.

Since the September 11 terrorist attacks, the federal government has spent large sums of money on facial recognition technology and donated millions of dollars to state and local governments to build their databases.

How does Facial Recognition Algorithms Work?

Face detection performs it first. Algorithms typically cycle through multiple frames looking for faces of a given size. Within these boxes, the system detects facial markings and assigns a score, providing a level of confidence that the image is a face. Once confirmed, face technology typically creates a template founded on factors such as the relative distance between the eyes and the point. And just below the nose, above the lip from ear to ear. The developed mathematical representation then compares to other detected faces.

The similarity in the proportions between the distances at various points on the face, typically centred around anchor points such as the nose, eyes, ears, and mouth, gives a score on the logarithmic scale. Close matches vary from 3 to 5, and apparent mismatches are less than 1. More than 40 points are possible when the same image serves as both research and target.

Benefits Of Facial Recognition Technology

Facial recognition technology benefits federal agencies, particularly law enforcement, defence, and intelligence agencies. However, it can also assist civic organizations in their humanitarian work.

1. Poor Image Quality Limits The Effectiveness of Face Detection

Image quality affects how to face recognition algorithms work. The image quality of video scanning is relatively low compared to a digital camera. Even high-definition video is 1080p at best (progressive scan), usually 720p. These values correspond to approximately 2 MP and 0.9 MP, respectively, while a cheap digital camera reaches 15 MP. The difference is quite apparent.

2. Small Image Sizes Make Face Recognition Difficult

When a face detection algorithm discovers a face in an image or still image without capturing a video, the relative size of that face compared to the size of the recorded image affects how well the face is recognized. An already small image size, combined with a remote camera subject, means that the noticed face is only 100 to 200 pixels on One side. Also, scanning an image for different face sizes is a processor-intensive activity. Most algorithms allow Specifying a range of face sizes to help eliminate false positive detection and speed up image processing.

3. Different Angles of the Face May Affect the Reliability of Face Recognition

The relative angle of the target’s face profoundly affects the recognition score. When registering a look in recognition software, multiple angles are often used (profile, front and 45 degrees are shared). Less than the front view affects the algorithm’s ability to generate a template for the face. The more direct and the higher the resolution of the image (both the recorded image and the research image), the higher the resulting match score.

4. Data Processing and Storage May Limit Facial Recognition Technology

Although high-definition video is relatively low resolution compared to digital camera images, it still takes up a significant amount of disk space. Processing every video frame is a huge task, so only a fraction (10 to 25 per cent) usually goes through a recognition system. Agencies can use pools of computers to minimize overall processing time. However, adding computers involves transferring a significant amount of data over a network, which limits it by input and output restrictions, further limiting processing speed.